While it's easy enough to delete rows from a manual index by @rid (using DELETE FROM INDEX:testindex WHERE rid = #10:4), is the index structured in a way that this is efficient? In other words, does ODB have to essentially read through the entire index to find all of the occurrences of @rid #10:4 to delete or is there any sort of optimization in place to prevent this? I'm concerned about using this approach on manual indexed that could have millions of records.
Also, I noticed that doing something like SELECT key FROM INDEX:textindex (WHERE...) doesn't work if the index has a composite key (works fine on simple keys). A bug or is there a different syntax that will work with composite keys?